There’s been a great deal of discussion of what we can learn from the Covid policy adopted by Sweden.  One side suggests that the Swedish outcome shows that lockdowns don’t have much impact on Covid infection rates, while the other side reaches the opposite conclusion.  I’m rather skeptical about the effectiveness of lockdown policies, but I don’t entirely agree with either side of the debate over Swedish policy.  (This article in The Economist is also mildly skeptical of the cost efficiency of lockdowns, accounting for both the impact on the economy and on human freedom.)  At the same time, I’ve long suspected that the Swedish government didn’t handle the epidemic very well.  That requires some explanation.

Scott Alexander has a very long and thoughtful article on the effectiveness of lockdowns, and does a good job of presenting both points of view.  Here’s what he says about Sweden:

Anyway, a reasonable conclusion might be that Sweden had between 2x (if we compare it to an average European country) and 6x (if we compare it to an average Scandinavian country) the expected death rate in the first phase of the pandemic.

Philippe Lemoine is extremely against this conclusion. He first argues that since we don’t know why Finland+Iceland+Norway+Denmark did so well, we can’t assume it’s a “Scandinavia effect” and so we can’t assume Sweden would share it. Therefore, we should be judging it against the European average rather than the (better) Scandinavian average. I would counter that, although we can’t prove that just because X is true of Finland+Iceland+Norway+Denmark and not other European countries, that it should also be true of Sweden, but we should have a pretty high prior on it . . .

That’s exactly my view.  In the past, I did a cross-sectional study of developed countries, and while doing so I found that the Nordic countries are actually quite distinctive.  According to a wide range of measures they are quite similar to each other and quite different from other countries.  So I have the same prior as Alexander, and thus lean toward the view that Sweden really did do much worse.  But . . . I don’t believe this was primarily due to Sweden’s lockdown policies.

Recall that Sweden is just a single data point.  When we look at more systematic studies, the effect of lockdowns seems to be much smaller.  Here is Alexander, citing a study of US states that found some impact from lockdowns:

This is a victory for lockdowns insofar as the correlation is significant, but strong proponents might be surprised by how small the effect was. A few small isolated northern states like Vermont did very well. But most states – from California [green dot] and New York to Florida [red dot] and Texas – clustered in a band between 80,000 and 120,000 cases per million. States at the 75th percentile of lockdown strictness had about 17.5% fewer cases per million than states at the 25th percentile.

And even that may overstate the effect, as people in states with stricter lockdown policies might also contain people that would be more careful in the absence of lockdowns.  On the other hand, Alexander points out that there was lots of voluntary behavioral change in states with weak lockdown policies, so this doesn’t suggest that we should have just lived life as normal during Covid, rather it casts some doubt on mandatory lockdowns having much marginal benefit.

So why did Sweden do far worse than its Nordic neighbors while places like South Dakota did only marginally worse than states with more restrictive lockdowns?  I don’t know, but I suspect it might have had something to do with other Swedish policies.  Here it’s useful to recall that Sweden has a high level of what’s sometimes called “civic virtue”, which is the tendency of the public to cooperate with what is seen as sound public policies.  Sweden is a “high trust society”.  And at least early on in the pandemic, some Swedish public officials seemed to be recommending a “herd immunity” approach to the pandemic.  Consider the following, from August 2020:

While most of the world has come to terms with covering their noses and mouths in crowded places, people in Sweden are going without, riding buses and metros, shopping for food, and going to school maskless, with only a few rare souls covering up.

Public  officials here argue that  are not effective enough at limiting the spread of the virus to warrant mass use, insisting it is more important to respect social distancing and handwashing recommendations.

“I think it’s a little bit strange. Sweden, as a small country, they think they know better than the rest of the world. (It’s) very strange,” says Jenny Ohlsson, owner of the Froken Sot shop selling colourful fabric masks in Stockholm’s trendy Sodermalm neighbourhood.

Last year, I recall reading that when Swedish people were interviewed, they cited government recommendations when explaining their lax behavior. In my view, Sweden ended up with much higher caseload than its Nordic neighbors mostly because of voluntary differences in behavior, partly attributable to dubious recommendations from the Swedish authorities.  That would explain why lockdowns seemed to have had a much more damaging effect in Sweden than elsewhere; the actual problem was a range of other behavioral differences that were not directly due to a lack of mandatory lockdowns.  (Some also cite Sweden’s inept nursing home management, but even that doesn’t fully explain the differences.)

So what are we left with?  Lockdowns are overrated in importance in two different ways.  First, they didn’t save nearly as many lives as their proponents assumed.  Second, their economic cost was lower than their opponents assumed.  Most of the economic damage was due to voluntary behavioral changes.  But lockdowns also restricted human freedom.

On balance, I lean toward the view that the gains in public health were not large enough to offset the admittedly modest economic damage, plus the substantial loss of freedom.  And this view is not based on any dogmatic opposition to any and all government regulation.  If the case fatality rate had been 60% instead of 0.6%, then lockdowns might have made sense as a temporary policy.  But even in that case it’s not entirely clear they would have been necessary, as the voluntary behavior response would have been far greater.  Rather the restriction on human freedom that would have made the most sense is a temporary restriction on inbound airline flights.

PS.  Lockdown policies had virtually zero impact on my life over the past 15 months, so my views here do not reflect grouchiness over being inconvenienced.  Rather, I believe that lockdown policies did restrict the behavior of many other people.

PPS.  This post does not apply to countries that successfully implemented a near-zero Covid policy, such as Australia and New Zealand.  It applies only to countries where Covid was widespread.  In a future pandemic, it is almost inevitable that all developed countries will try to emulate Australia, at least initially.  You may not like that fact, but it seems inevitable to me.

PPPS.  This issue is almost endlessly complicated, as Sweden did have some “lockdown” rules, and later in the epidemic it made the rules more restrictive in response to soaring caseloads.  As a result, Scott Alexander focused mostly on the early phase of the epidemic, to minimize the “endogeneity problem”.  Similarly, “herd immunity” was never the official Swedish policy, rather government officials spoke of it as a reasonable objective.